Cloud Cost Optimization, GCP

Commitment-Based Discounts Across AWS, GCP and Azure

How commitment pricing works, how the two commitment types differ, and how AWS, GCP and Azure each structure them.

Aman Singh

Technical Content Writer, FinOps

Updated August 26, 2026

17 min read

Key takeaways

Ceilings run up to 72% on AWS and Azure and up to 70% on GCP, but those are maximums tied to specific machine series and three-year terms. What you realise depends on which commitment type you buy and how much of it you use.

An unused commitment still bills at the committed rate for the full term. Exit routes are narrow: Standard RIs can be resold, Azure reservations exchanged or refunded within limits, GCP CUDs neither.

Commit only to the base load you can forecast, and leave variable demand on on-demand or spot. Provider docs frame sizing as a break-even calculation, not a fixed coverage target.

Commitment-based discounts are pricing agreements with AWS, GCP, and Azure. You commit to a set amount of hourly spend or eligible resource usage for a one- or three-year term in return for discounted rates. This guide focuses primarily on compute-related commitment products; discount ceilings and eligibility vary by provider, service, configuration, term, payment option, and utilization.

The Short Answer

The commitment products differ by name across AWS, GCP and Azure, but they resolve to the same two choices: commit to a level of spend and stay flexible, or commit to a specific resource and take the deeper rate.

Which one fits depends on whether your workloads hold their shape for the length of the term, not on which cloud you run.

That is a starting point, not a rule. Discount ceilings, term options and exit routes still differ by provider, and a commitment bought on one cloud never carries over to another.

Spend-Based vs Resource-Based Commitments

Spend-based commitments, including AWS Compute Savings Plans, GCP compute flexible CUDs, and Azure savings plan for compute, commit you to a minimum hourly spend.

They apply automatically to eligible usage within each product’s service, account, and scope rules; AWS EC2 Instance Savings Plans are a narrower spend commitment limited to one instance family in one Region.

Resource-based commitments, including Reserved Instances, GCP resource-based CUDs, and Azure Reserved VM Instances, generally require a more specific resource, machine family, or regional configuration.

They can offer deeper rates, but eligibility can fall when the workload changes outside the commitment’s applicable configuration.

	Spend-based	Resource-based
Workload stability	Shifting families or regions	Stable configuration
Portability	Varies by product; broad eligible usage*	Fixed to the committed resource
Discount depth	Usually lower ceiling*	Deepest available
You must forecast	A spend level	The exact resource shape
Best for	Evolving estates	Steady-state baselines

*Spend-based flexibility and discount depth are product-specific, not universal. Most spend-based commitments apply across instance families and regions, but AWS EC2 Instance Savings Plans apply only to a single instance family in one Region, and reach the same up-to-72% ceiling as Standard RIs.

How Commitment-Based Discounts Work

You choose a commitment level: an hourly spend (for example, $10/hour) or a resource quantity and a one- or three-year term.
Discounts apply automatically to matching usage. No infrastructure or code changes are required.
Usage above your committed amount is billed at standard on-demand rates.
Unused commitment still bills at the committed rate. It does not roll over.

Three variables set the discount depth:

Term. Three-year commitments earn deeper rates than one-year.
Payment. The more you pay upfront, the deeper the rate where upfront options exist; GCP bills commitments monthly only.
Specificity. The more precisely defined the commitment, the deeper the rate, and the harder it is to keep fully utilized.

Two metrics measure the result. Coverage is the share of eligible spend covered by commitments; utilization is the share of purchased commitment actually consumed.

Read them together with high utilization at low coverage means you are under-committed, as Google defines both metrics.

For sizing, Google’s CUD recommendation guidance frames it as break-even: subtract the discount percentage from 100 to estimate the utilization at which a commitment stops paying off.

AWS’s pricing-model guidance advises against fixed coverage targets, buy small amounts incrementally and keep purchasing while estimated savings stay above your threshold.

Coverage

eligible spend covered by commitments ÷ total eligible spend

Utilization

commitment consumed ÷ commitment purchased

Break-even utilization

100 − discount %

Read both together — high utilization at low coverage means you are under-committed.

Commitment Discount Types by Provider

Every provider implements both structures under its own names and a commitment purchased on one cloud never applies to another.

	AWS	GCP	Azure
Max discount	Up to 72%	Up to 70%*	Up to 72%
Term options	1 or 3 years	1 or 3 years; eligible resource-based (hardware) CUDs can extend to custom terms†	1 or 3 years
Payment options	All, partial, or no upfront	Monthly only	All upfront or monthly
Resale or transfer	RI Marketplace (Standard RIs only)	None	Exchange or refund, within limits

*GCP’s maximum applies to 3-year resource-based CUDs on memory-optimized machine series; other machine series reach up to 55%, and compute flexible CUD rates run 17–63% by machine series and term.

Term extensions apply only to eligible Compute Engine hardware (resource-based) commitments, not to compute flexible commitments.

A 1-year commitment can extend to a custom term longer than 1 year and shorter than 3; a 3-year commitment to a custom term longer than 3 years and shorter than 6. See Google Cloud’s guidance on extending commitment terms.

Maximum discounts vary by eligible service, machine family, region, term, payment option, operating system, and utilization. Check the current provider pricing page before purchase.

AWS: Savings Plans and Reserved Instances

AWS offers four Savings Plans types: Compute Savings Plans (up to 66% across EC2, Fargate, and Lambda), EC2 Instance Savings Plans (up to 72%, locked to one instance family in one Region), Database Savings Plans (up to 35% across RDS, Aurora, and other database services), and SageMaker AI Savings Plans.

Reserved Instances run alongside them: Standard RIs reach up to 72%, Convertible RIs up to 66% with the flexibility to exchange attributes mid-term. Full breakdowns in our guides to AWS Savings Plans and AWS Reserved Instances.

GCP: Committed Use Discounts

GCP structures both models as Committed Use Discounts: resource-based CUDs commit to vCPU and memory in a region (up to 55%, or up to 70% for memory-optimized series), while compute flexible CUDs commit to hourly spend across Compute Engine, GKE, and Cloud Run (28% and 46% for most general-purpose series).

GCP is also the only provider with an automatic no-commitment tier, sustained use discounts compared in our SUD vs CUD guide. Full rates and sizing in our GCP CUD guide.

Azure: Reservations and Savings Plans

Azure Reserved VM Instances commit to a VM family and region for up to 72%, while Azure savings plan for compute commits to hourly spend across VM families and regions for up to 65%.

When both apply to the same VM, Azure applies the reservation first. Full breakdowns in our guides to Azure reservations and Azure savings plans, and our side-by-side comparison.

Commitment vs On-Demand vs Spot

On-demand. Full price, full flexibility. Best for unpredictable, short-lived, or new workloads.
Commitments. Discounted rates for the stable base load you can forecast across the term.
Spot. Up to 90% off for interruptible capacity. Best for fault-tolerant batch and CI workloads spot usage is already discounted and sits outside commitment coverage.

Mature teams layer all three: commitments under the base load, on-demand for variable demand, spot for interruptible peaks.

Benefits of Commitment-Based Discounts

Lower unit costs. Matching workloads run at provider-discounted rates, but realized savings depend on the product, term, payment option, configuration, and commitment utilization.
Budget predictability. Committed spend is fixed for the term, giving finance teams a firmer baseline across planning cycles.
No engineering effort. The discount is a billing construct with nothing about the infrastructure changes.

Risks FinOps Teams Must Manage

Underutilization. An unused commitment bills at the committed rate whether or not matching workloads run. A commitment purchased on last year’s usage patterns can become a stranded cost when architectures change mid-term.
Lock-in. Most commitments cannot be cancelled once purchased, so every purchase is a forecast. Over-commit and you pay for capacity you don’t use; under-commit and you leave savings on the table.
Multi-cloud complexity. Each provider defines coverage and utilization differently, runs its own renewal calendar, and offers different exit routes and no commitment ports across clouds. Managing three portfolios is three times the tracking, not one consolidated view.

Commitment Management Maturity

Commitment capability typically matures through three stages:

Ad hoc. Individual purchases made from provider console recommendations, reviewed rarely if at all.
Systematic. Baseline analysis on the providers’ recommendation look-back windows (7–60 days), workloads segmented by stability before purchase, and a fixed review cadence.
Continuous. Automated monitoring of coverage and utilization, incremental purchases as usage grows, and renewals re-sized against current usage rather than repeated.

At every stage the same rule holds: commit to the base load you can defend with data, and let variable demand stay on on-demand or spot.

How Usage.ai Fits

The last of those stages, continuous automated optimization, is where we work. We analyse billing and usage data across AWS, Azure, and GCP and generate updated commitment recommendations every 24 hours, so coverage decisions track how workloads behave now rather than how they behaved last quarter.

With Flex Insured Commitments, teams can get the 30–50% savings of a 1- or 3-year commitment with none of the commitment risk. That applies across covered workloads on AWS, Azure, and GCP, and after a customer approves a recommendation, we purchase and manage the commitment on their behalf.

If a commitment ever costs more than the equivalent on-demand usage would have, we calculate that loss at the end of the month and provide cashback protection for it. We charge a percentage of realized savings, so the fee only exists when the savings do.

Conclusion

Commitment-based discounts are the largest pricing lever available on AWS, GCP, and Azure and the only one that carries term risk. Size against the base load you can forecast, and treat every renewal as a fresh decision rather than a repeat purchase.

EVALUATE YOUR COMMITMENT COVERAGE

See what your commitments cover.

Review base load, uncovered spend, and commitment risk across clouds.

Book a Savings Assessment

Frequently asked questions

How do commitment-based discounts compare to on-demand pricing?

On-demand is the list price with no obligation. For the compute-focused commitment products covered here, commitments trade a one- or three-year term for rates up to 72% below it on AWS and Azure and up to 70% on GCP. Realized savings depend on the product, term, payment option, configuration, and commitment utilization, and the commitment bills whether you use it or not.

Are commitment-based discounts worth it for smaller teams?

Yes, once there is a stable usage baseline to commit against. Start with a small commitment covering the portion of spend that has held steady, and keep genuinely variable workloads on on-demand until a pattern emerges.

What happens if we over-commit or need to exit?

AWS: Standard RIs can be resold on the Amazon EC2 Reserved Instance Marketplace; Savings Plans cannot be sold or cancelled.
Azure: Reservations can be exchanged subject to product and value rules, or refunded within a $50,000 rolling 12-month cancellation cap. From February 1, 2027, reservations purchased on or after that date cannot be exchanged when the corresponding service is supported by savings plans; reservations purchased before that date retain one final exchange after that date.
GCP: CUDs cannot be cancelled, exchanged, or resold once purchased.

Can commitment discounts stack?

Within one provider, yes, they apply in a fixed order rather than doubling up. AWS applies Reserved Instances before Savings Plans, Azure applies reservations before savings plans, and GCP applies resource-based CUDs before flexible CUDs.

https://learn.microsoft.com/en-us/azure/cost-management-billing/reservations/reservation-exchange-policy-changes